Model Overview

The djuna/MN-Chinofun-12B-2 is a 12 billion parameter language model developed by djuna. It was created using the Model Stock merge method, leveraging ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2 as its base model. This merging approach combines the strengths of multiple pre-trained models to enhance overall performance and versatility.

Merged Components

This model integrates five distinct pre-trained models:

• grimjim/magnum-consolidatum-v1-12b
• spow12/ChatWaifu_v1.4
• GalrionSoftworks/Canidori-12B-v1
• Nohobby/MN-12B-Siskin-v0.2
• RozGrov/NemoDori-v0.2.2-12B-MN-ties

Performance Metrics

Evaluations on the Open LLM Leaderboard indicate an average score of 25.37. Specific scores include:

• IFEval (0-Shot): 61.71
• BBH (3-Shot): 29.53
• MATH Lvl 5 (4-Shot): 11.18
• MMLU-PRO (5-shot): 29.06

These results provide insight into its capabilities across various reasoning and knowledge-based tasks. The model supports a context length of 32768 tokens.